The three-dimensional surface reconstruction and high-precision volume calculation of spatially irregular objects, such as above-ground buildings, slopes, and underground goafs, hold significant application value in modern logistics, warehousing, and mine surveying. To address the problems of noise interference, non-uniform density distribution, and inaccurate normal vector estimation in complex curved regions of laser point clouds acquired under real-world conditions, a series of optimization strategies are proposed to improve three-dimensional modeling quality and volume calculation accuracy. First, an adaptive feature-preserving filtering algorithm is proposed by introducing local density and curvature features into the conventional statistical filtering framework, which effectively suppresses outlier noise while significantly reducing the misclassification and removal of feature points in sparse regions and high-curvature areas. Second, to overcome the degradation of normal vector estimation accuracy caused by non-uniform point cloud density, a curvature-adaptive hybrid normal estimation strategy is developed. Principal component analysis(PCA) is employed for rapid normal estimation in relatively flat regions, whereas quadratic surface fitting is introduced for local refinement in regions with significant curvature variation, thereby improving the stability and accuracy of normal estimation in complex surface regions while maintaining computational efficiency. Finally, watertight surface reconstruction is achieved using a smooth signed distance implicit function, and volume calculation is performed via a projection-tetrahedral decomposition method. The proposed method is validated through multi-scenario experiments involving a bucket model, a laboratory space, and a laboratory-corridor combined structure. The resulting relative volume errors are reduced to 1.63%, 0.32%, and 0.25%, respectively. Experimental results demonstrate that the proposed method achieves improved surface reconstruction continuity and volume calculation accuracy, providing a stable and reliable technical solution for the high-precision three-dimensional surface reconstruction and volume measurement of complex spatial objects.